id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
2d5db80b-1cbb-4705-b7e8-03cd8b7e72da | trentmkelly/LessWrong-43k | LessWrong | Alternative Places to Get Ideas (Also, "In Defense of Food")
I discovered Less Wrong a few months ago (courtesy of Harry Potter and the Methods of Rationality), and I am extremely grateful to have such a thoughtful place to have discussions and to learn new things. But one of the more significant hazards I've become a... |
f40dc3bd-48a7-478e-9469-8b2777e08576 | trentmkelly/LessWrong-43k | LessWrong | Toki pona FAQ
Whenever I start telling someone about toki pona, they ask at least some of these questions. So I compile the questions and my answers here.
Toki pona is a constructed language notable for having under 200 words. The strange writing that probably prompted you to ask me about it is sitelen pona.
How do... |
af193a53-8ca5-4f16-8bb5-cdfba46d7029 | trentmkelly/LessWrong-43k | LessWrong | Education via Cell Phone in Africa
The World Bank reports that between 2007 and 2012, Africa exceeded the number of cell phone subscribers found in the European Union and in the United States. Africa has 650 million cell phone subscribers. [1] At the same time, Africa has profound infrastructure gaps that the World ... |
8588c714-4ffe-4c6c-b40b-beea9ac7c03b | trentmkelly/LessWrong-43k | LessWrong | Is it true that only a chatbot encouraged a man to commit suicide?
Background: this article. TL;DR: A man who talked with a chatbot about climate change for weeks was recommended by it to commit suicide.
This story got into several Belgian newspapers, but IIRC never made it to the state media, somewhat to my surpris... |
1df7b599-8694-4316-bffb-6cb3e42451d6 | trentmkelly/LessWrong-43k | LessWrong | Which ML skills are useful for finding a new AIS research agenda?
TL;DR: I'm trying to either come up with a new promising AIS direction or decide (based on my inside view and not based on trust) that I strongly believe in one of the existing proposals. Is there some ML background that I better get? (and if possible: ... |
646e6916-73f0-470d-9d34-ff61052b8998 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Troubling Trends in Machine Learning Scholarship
1 Introduction
---------------
Collectively, machine learning (ML) researchers
are engaged in the creation and dissemination of knowledge about data-driven algorithms.
In a given paper, researchers might aspire to any subset of the following goals, among others:
to t... |
4f769b75-f2b1-40a1-aa5c-0827ca942ef4 | trentmkelly/LessWrong-43k | LessWrong | Missives from China
Driving in China
You’ve heard of the Iterated Prisoner’s Dilemma. But have you heard of Iterated Chicken?
Heat Lag
Day 4: Temperature is experienced differently here. People wear wool coats over turtlenecks in 70 degree weather. Counted two other people wearing T-shirts out of thousands.
Day 5: ... |
303d11a0-691b-4777-a95e-d20780663c67 | trentmkelly/LessWrong-43k | LessWrong | Security analysis of 'cloud chemistry labs'?
I learned of 'cloud chemistry labs' while watching this video:
* Stephen Wolfram: Complexity and the Fabric of Reality | Lex Fridman Podcast #234 - YouTube
Wolfram is trying to devise a 'molecular computing' experiment to generate prime numbers using the 'dynamics' of a ... |
cea295d4-63e6-4001-89bc-47ffa020c116 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Ramble on STUFF: intelligence, simulation, AI, doom, default mode, the usual
Cross posted from [New Savanna](https://new-savanna.blogspot.com/). I've published a [number of these rambles](https://new-savanna.blogspot.com/search/label/Ramble) over at New Savanna. They are a way for me to think about what's on my mind a... |
34dc91f5-db68-4b05-b096-168bbc1b8861 | trentmkelly/LessWrong-43k | LessWrong | [Meta] No LessWrong Blackout?
Our sister site apparently is:
> Overcoming Bias will resume normal service on Jan 18 8:00pm (Eastern Time).
> Today Overcoming Bias joins Wikipedia and many other sites in protesting proposed legislation — the Stop Online Piracy Act (SOPA) in the U.S. House of Representatives, and the ... |
9476eb64-8719-45b2-be93-c611f08ec855 | trentmkelly/LessWrong-43k | LessWrong | Criticisms of the Metaethics
I'll admit that I'm using the LessWrong Board to try and figure out flaws in my own philosophical ideas. I should also make a disclaimer that I do not dispute the usefulness of Eliezer's ideas for the purposes of building a Friendy AI.
My criticisms are designed for other purposes- namely... |
5d6f726d-f63f-4512-99fc-69999850268e | trentmkelly/LessWrong-43k | LessWrong | The Sandwich Argument
Today I saw a very interesting argument pattern.
> Normally I'd expect some evidence for X. I don't think there was any evidence for X apart from Y. So in conclusion, there was no evidence for X.
It's well known that people tend to remember the start and end more than the middle. This pattern i... |
ef2fb600-9591-4a04-b5d8-3d8d1d10b3d8 | trentmkelly/LessWrong-43k | LessWrong | SimpleStories: A Better Synthetic Dataset and Tiny Models for Interpretability
The dataset and model suite TinyStories is widely used as a toy setup for mech interp research. Searching for the term on this forum currently yields 23 results. "A List of 45+ Mech Interp Project Ideas from Apollo Research’s Interpretabili... |
9ca5a118-e77a-4b5a-a754-d0b58572dd8c | trentmkelly/LessWrong-43k | LessWrong | Meetup : DC Meetup: NoVa
Discussion article for the meetup : DC Meetup: NoVa
WHEN: 07 September 2011 07:00:00PM (-0400)
WHERE: Ballston Commons Mall
We're going to talk about Non-Violent Communication, come prepared with questions, and some willingness to practice.
Discussion article for the meetup : DC Meetup: N... |
456eee30-581f-4586-bcbb-75a8a293d58e | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Documenting Journey Into AI Safety
**Summary**
-----------
Recently, I have noticed that many individuals have become discouraged due to a seeming lack of progress in their attempts to break into a career in AI safety. It seems to me that a key bottleneck in the *beginning a career in AI safety* pipeline is access to... |
a90ca213-0a1f-48ea-a2d6-231e4d58e5f8 | trentmkelly/LessWrong-43k | LessWrong | Book Review: Kazdin's The Everyday Parenting Toolkit
Follow-up to Rational Parenting, Ignorance in Parenting, Soft Paternalism in Parenting
Part of a Sequence on Rational Parenting
This is a review of The Everyday Parenting Toolkit: The Kazdin Method for Easy, Step-by-step, Lasting Change for You and Your Child by Al... |
2955c35c-86ee-4890-914f-d3d00c00b20a | trentmkelly/LessWrong-43k | LessWrong | Infinite Summations: A Rationality Litmus Test
You may have seen that Numberphile video that circulated the social media world a few years ago. It showed the 'astounding' mathematical result:
1+2+3+4+5+… = -1/12
(quote: "the answer to this sum is, remarkably, minus a twelfth")
Then they tell you that this result is... |
68194c05-6812-4b9b-9f34-0e4587a498d7 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | [LINK] OpenAI doing an AMA today
The [OpenAI](/openai.com) research team is doing a [Reddit AMA](https://www.reddit.com/r/MachineLearning/comments/404r9m/ama_the_openai_research_team/) today! A good opportunity to ask them questions about AI safety and machine learning. |
ff96d2f0-07d1-4f73-9fa2-a112a1983110 | trentmkelly/LessWrong-43k | LessWrong | A Hill of Validity in Defense of Meaning
> If you are silent about your pain, they'll kill you and say you enjoyed it.
>
> —Zora Neale Hurston
Recapping my Whole Dumb Story so far—in a previous post, "Sexual Dimorphism in Yudkowsky's Sequences, in Relation to My Gender Problems", I told the part about how I've "alwa... |
542594ea-dbc6-4b48-96a2-6e8b7b87dca9 | trentmkelly/LessWrong-43k | LessWrong | On knowing the Future
Across peaks of inspired productive days, I get lost inbetween valleys of foggy unfocused wandering.
Some years ago, I begin recording the daily fluctuations of concrete output and work hours, which unsurprisingly maps to the Hand-Eye Dilemma that one would find in the creative rhythms of artist... |
61585457-928d-430e-9be5-bb7a448a586f | trentmkelly/LessWrong-43k | LessWrong | Does scientific productivity correlate with IQ?
Anders Ericsson, in his popular book Peak: Secrets from the New Science of Expertise, claims that though one may need a reasonably high IQ to even be a scientist, after that level, IQ is irrelevant to scientific success.
> The average IQ of scientists is certainly highe... |
c31c2363-17f8-4a1d-be4a-c35c691886e8 | StampyAI/alignment-research-dataset/blogs | Blogs | Reflections on the first year of parenting
The first year after having a baby went by really fast – happy birthday Daniel! This post is a reflection on our experience and what we learned in the first year.
[](https://... |
c84fec20-3914-42a9-a6e9-33e929a5edfd | trentmkelly/LessWrong-43k | LessWrong | Why Aren't There More Schelling Holidays?
A common pattern when working in teams: when one person is out on vacation, there’s a disproportionate drop in productivity for the team as a whole. Lots of things end up blocked on the person who’s out. For instance, in a small software team, maybe the developer who owns a pa... |
410e2404-1460-4e96-b1a3-926177625695 | trentmkelly/LessWrong-43k | LessWrong | (notes on) Policy Desiderata for Superintelligent AI: A Vector Field Approach
Meta: I thought I'd spend a little time reading the policy papers that Nick Bostrom has written. I made notes as I went along, so I spent a little while cleaning them up into a summary post. These are my notes on Bostrom, Dafoe and Flynn's 2... |
405b554d-8759-462f-9432-5267436c1bb9 | StampyAI/alignment-research-dataset/arxiv | Arxiv | User Tampering in Reinforcement Learning Recommender Systems
1. Introduction
----------------
*Recommender systems* are algorithms that filter through large pools of some entity in the hope of identifying and recommending particularly relevant entities of interest to an individual or group (Bobadilla et al., [2013]... |
e40a9dad-4785-494c-a4e6-a9b8ec860b14 | StampyAI/alignment-research-dataset/arbital | Arbital | Addition of rational numbers exercises
This page consists of exercises designed to help you get to grips with the addition of rational numbers.
# Core exercises
## First exercise: $\frac{1}{10} + \frac{1}{5}$
%%hidden(Show some possible solutions):
Using the instant rule from the text (which is actually a bit unwie... |
fcec97e8-6612-4b53-baee-12e449da9f18 | trentmkelly/LessWrong-43k | LessWrong | Sandboxing by Physical Simulation?
This is a simple idea. I do not remember if I've seen it anywhere. It is probably not original, but I am mildly surprised that I haven't come across it, even if only to see it refuted. If this is an old/dumb idea, please let me know and I'll delete it.
People have built universal Tu... |
6ccc59ff-e554-4d80-9b6d-14bd04f7ee86 | trentmkelly/LessWrong-43k | LessWrong | Quotes Worth Talking About
Why do we need quotes? The process of synthesizing information and distilling it into the most accessible form possible is generally useful for disseminating information. I believe that one of the general goals of this board is to increase awareness in topics of rationality, AI alignment the... |
ce8eda3c-ae06-4573-b818-d4de4be12aef | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | [Link post] How plausible are AI Takeover scenarios?
*This is a linkpost for two posts written by myself and* [*Sam Clarke*](https://www.lesswrong.com/users/sam-clarke)*,* [***Distinguishing AI Takeover Scenarios***](https://www.lesswrong.com/posts/qYzqDtoQaZ3eDDyxa/distinguishing-ai-takeover-scenarios)*and* [***Inves... |
76d830dc-4144-4b53-8753-8de3467d6a1f | trentmkelly/LessWrong-43k | LessWrong | Darknet Mining for Proactive Cybersecurity Threat Intelligence
They are using machine learning to comb the darknets, capturing about 300 threats a week.
About 90% hack application and backdoor recognition, that is for sale, and about 80% hacker forum vulnerability identification.
"These threat warnings include infor... |
d7dbea09-c1ec-4da9-aa75-3ad1b8c8ba96 | trentmkelly/LessWrong-43k | LessWrong | Empirical Sleep Time
I'm thinking that it should be possible to decide when to sleep based on reduced performance.
Can anyone suggest a tool for that purpose? Perhaps some reaction time testing software?
I guess I would have to track myself during the day to make a baseline, which is fine.
But without some sort of... |
6361489d-5174-4c62-8097-9cf2dd9c1915 | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups: Austin, Berlin, Melbourne, NYC, Paris, Pittsburgh, Washington DC
This summary was posted to LW Main on August 31st, and has been moved to discussion.
There are upcoming irregularly scheduled Less Wrong meetups in:
* Paris Meetup: 01 September 2012 02:00PM
* Washigton DC (probably) minicamp meetu... |
c305ff61-f4a3-4904-890a-fa8e19dcd29b | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | EIS XI: Moving Forward
Part 11 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7).
So far, this sequence has discussed a number of topics in interpretability research, all building toward this post. Its goal is to explain some approaches that may be valuable moving... |
e96caf25-5a00-4be9-b801-c490e8737f03 | trentmkelly/LessWrong-43k | LessWrong | Who determines whether an alignment proposal is the definitive alignment solution?
What is the established process, or what could be a potential process, for validating a highly probable alignment solution? I assume that a robust alignment solution would undergo some form of review—so who is responsible for reviewing ... |
d62cac40-5c6a-4475-b8a7-7714a3b0afd3 | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | What links are especially valuable to share on social media or other contexts?
Sharing high quality information about AI Safety can be one of the lowest-effort ways to expose people to the ideas. Be sure to engage with the replies with care and do your research when replying to questions (feel free to add such questio... |
f4fd9467-7bb8-4c2d-bbc3-f33bfeb99a83 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Playing with DALL·E 2
I got access to [Dall·E 2](https://openai.com/dall-e-2/) yesterday. Here are some pretty pictures!
My goal was to try to understand what things DE2 could do well, and what things it had trouble understanding or generating. My general hypothesis is that it would do a better job with things that a... |
89f97b73-ca54-459d-b92f-5b03addcf537 | trentmkelly/LessWrong-43k | LessWrong | The Median Researcher Problem
Claim: memeticity in a scientific field is mostly determined, not by the most competent researchers in the field, but instead by roughly-median researchers. We’ll call this the “median researcher problem”.
Prototypical example: imagine a scientific field in which the large majority of pr... |
73472170-416b-44f0-9020-79bf8b28b9dc | trentmkelly/LessWrong-43k | LessWrong | Satisf-AI: A Route to Reducing Risks From AI
(Ok so I admit the word “Satisf-AI” is weird, I was intending for it to be a play off of “Satisfy” and “AI” and for it to be pronounced like “Satisfy”)
I think Holden Karenofsky’s PASTA is a good way to frame the discussion of transformative AI. It resonates with something... |
a4e70887-65a6-437d-8370-96c59b05a2ae | trentmkelly/LessWrong-43k | LessWrong | 14 Techniques to Accelerate Your Learning
Originally written by Belén Cobeta and Spencer Greenberg from ClearerThinking.org. Revised by Teis Rasmussen & Florence Hinder from ThoughtSaver.com. This is a cross-post from the Effective Altruism Forum.
TLDR: There are a number of techniques that may accelerate the speed a... |
e826dfbc-39e2-42c7-b45c-cfd963337c68 | trentmkelly/LessWrong-43k | LessWrong |
[MLSN #3]: NeurIPS Safety Paper Roundup
As part of a larger community building effort, I am writing a safety newsletter which is designed to cover empirical safety research and be palatable to the broader machine learning research community. You can subscribe here or follow the newsletter on twitter here.
--------... |
22451bb6-5bfe-4c4d-9252-0a23c6b2bff6 | trentmkelly/LessWrong-43k | LessWrong | Intervening in the Residual Stream
This post describes work done at SERI-MATS with mentorship from Neel Nanda. This work has benefitted from discussions with Neel Nanda, Stefan Heimersheim, Kajetan Janiak, Wes Gurnee, Marius Hobbhahn, Aryan Bhatt, Mykyta Baliesnyi, and Laura Vaughn.
Executive Summary
* We demonstra... |
de1f5800-cfc1-47f4-8d10-553fca39d2cc | trentmkelly/LessWrong-43k | LessWrong | Modularity and Buzzy
This is the second part in a mini-sequence presenting material from Robert Kurzban's excellent book Why Everyone (Else) Is a Hypocrite: Evolution and the Modular Mind.
Chapter 2: Evolution and the Fragmented Brain. Braitenberg's Vehicles are thought experiments that use Matchbox car-like vehicles... |
5b55dae8-ed6b-4bfa-98ba-555ef7123ab8 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Pivotal outcomes and pivotal processes
***tl;dr:** If you think humanity is on a dangerous path, and needs to "pivot" toward a different future in order to achieve safety, consider how such a pivot could be achieved by multiple acts across multiple persons and institutions, rather than a single act. Engaging more act... |
30dd2569-4085-4fb1-89b9-8348e1df7a13 | trentmkelly/LessWrong-43k | LessWrong | Practicing what you preach
LessWrongers as a group are often accused of talking about rationality without putting it into practice (for an elaborated discussion of this see Self-Improvement or Shiny Distraction: Why Less Wrong is anti-Instrumental Rationality). This behavior is particularly insidious because it is sel... |
591b53b0-1583-486a-a012-965f0e35ff24 | StampyAI/alignment-research-dataset/arbital | Arbital | Conditional probability
summary: $\mathbb{P}(X\mid Y)$ means "The [probability](https://arbital.com/p/1rf) that X is true, assuming Y is true."
- $\mathbb{P}(yellow\mid banana)$ is "the chance that something is yellow, given that we know it is a banana" or equivalently "the chance that a banana is yellow".
- $\mathb... |
f671cc7d-bad6-4842-bf1d-344a9e8edf18 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | My AI Alignment Research Agenda and Threat Model, right now (May 2023)
TLDR
----
Fairly short timelines, mildly fast takeoffs, and medium-high uncertainties. --> Looking for abstractions to help cognition-steering and value-loading. --> Grasping at / reacting to related/scary/[FOMO](https://en.wikipedia.org/wiki/Fea... |
afe3070f-1f67-437d-8b5b-e81af31b26fb | trentmkelly/LessWrong-43k | LessWrong | D&D.Sci Dungeonbuilding: the Dungeon Tournament Evaluation & Ruleset
This is a follow-up to last week month year's D&D.Sci scenario: if you intend to play that, and haven't done so yet, you should do so now before spoiling yourself.
There is a web interactive here you can use to test your answer, and generation code ... |
78771040-b805-470b-a20a-9d98169a3d59 | trentmkelly/LessWrong-43k | LessWrong | Feedback is central to agency
An agent seems to be a particular kind of device that creates a feedback loop between the world and itself. It seems to have a lot to do with devices whose behavior depends on the consequences of their own behavior: that is, devices that behave in ways that cause the overall system in whi... |
bb266dbe-9ce7-4450-b977-338a9f5fb2d8 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post681
There have been several discussions about the importance of adversarial robustness for scalable oversight. I’d like to point out that adversarial robustness is also important under a different threat model: catastrophic misuse . For a brief summary of the argument: Misuse could lead to catastrophe. AI-assi... |
f4b97f07-60c6-4d2c-8662-3a38ad53296d | trentmkelly/LessWrong-43k | LessWrong | Rationalists are missing a core piece for agent-like structure (energy vs information overload)
The agent-like structure problem is a question about how agents in the world are structured. I think rationalists generally have an intuition that the answer looks something like the following:
* We assume the world follo... |
03f673a6-8f59-424f-8165-2dc6a1145e85 | trentmkelly/LessWrong-43k | LessWrong | Grinding slimes in the dungeon of AI alignment research
This post aims to describe some differing views on challenges in AI alignment research (as I understand them) using an analogy to a video game with multiple levels.
I then draw some conclusions about the potential upsides and downsides of doing and publishing di... |
bbf0a0b0-8326-4bef-9228-d6041a6c2262 | trentmkelly/LessWrong-43k | LessWrong | Morality is not about willpower
Most people believe the way to lose weight is through willpower. My successful experience losing weight is that this is not the case. You will lose weight if you want to, meaning you effectively believe0 that the utility you will gain from losing weight, even time-discounted, will out... |
62c85303-c5db-4976-9cde-87293d02e409 | trentmkelly/LessWrong-43k | LessWrong | Tabletop Role Playing Game or interactive stories for my daughter
Hey,
my daughter is about 3 and a half year old and we have been enjoying some interactive story telling where she is manipulating the story somehow.
When I was ~18, I played tabletop RPG with my friends (for about a year). I really enjoyed it and I t... |
56c6f782-f2f0-45f7-a5a1-4b82b2ccce54 | trentmkelly/LessWrong-43k | LessWrong | Machine Unlearning in Large Language Models: A Comprehensive Survey with Empirical Insights from the Qwen 1.5 1.8B Model
By Saketh Baddam And Saahir Vazirani
Abstract
Machine unlearning—the targeted removal of harmful or private knowledge from large language models (LLMs)—has emerged as a cornerstone of responsible ... |
d8b40209-96d1-4a5f-ab55-efe825e39bc9 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Vael Gates: Risks from Highly-Capable AI (March 2023)
I gave a [talk introducing AI alignment / risks from advanced AI](https://forum.effectivealtruism.org/posts/q49obZkQujkYmnFWY/vael-gates-risks-from-advanced-ai-june-2022) in June 2022, aimed at a generally technical audience. However, given how fast AI has been mov... |
be1cbcaf-f0f1-47f3-a687-466342230ba5 | trentmkelly/LessWrong-43k | LessWrong | Monthly Roundup #4: March 2023
Before I begin, I have an announcement to make: My game, the Emergents Trading Card Game, has launched!
You can download the game here for free for Windows, Mac and Android. On Android, you’ll have to side load it for now.
While I am obviously biased: I highly encourage you to download... |
4b826f35-a75f-47cd-b0d5-7b262faa0290 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Mode collapse in RL may be fueled by the update equation
**TL;DR:** We present an advantage variant which, in certain settings, does not train an optimal policy, but instead uses a fixed reward to update a policy a fixed amount from initialization. Non-tabular empirical results seem mixed: The policy doesn't mode-coll... |
3ee57440-7303-4773-9bb0-ee629ad018c1 | StampyAI/alignment-research-dataset/blogs | Blogs | Import AI 321: Open source GPT3; giving away democracy to AGI companies; GPT-4 is a political artifact
Welcome to Import AI, a newsletter about AI research. Import AI runs on lattes, ramen, and feedback from readers. If you’d like to support this (and comment on posts!) please subscribe.
[Subscribe now](https://impo... |
5faa7953-7b33-4172-ab98-2d06942bb944 | trentmkelly/LessWrong-43k | LessWrong | Meetup : LessWrong Megameetup
Discussion article for the meetup : LessWrong Megameetup
WHEN: 04 February 2012 12:00:00PM (-0500)
WHERE: Leverage House, New York City (contact for address)
We're holding the second East Coast Megameetup this weekend at the Leverage House in NYC, kindly provided by Geoff and the Levera... |
a6cb2c6d-2685-4d6b-b777-4460d4a7c564 | trentmkelly/LessWrong-43k | LessWrong | [Link] The Dominant Life Form In the Cosmos Is Probably Superintelligent Robots
An Article on Motherboard reports about Alien Minds by Susan Schneider who claiThe Dominant Life Form In the Cosmos Is Probably Superintelligent Robots. The article is crosslinked to other posts about superintelligence and at the end disc... |
14d93c52-53ab-4ba3-9842-eeaded015cd2 | trentmkelly/LessWrong-43k | LessWrong | AI safety and the security mindset: user interface design, red-teams, formal verification
[This is part 2 of a 5 part sequence on how security and cryptography can help AI safety, published and linked here a few days apart.]
Parallels between the AI safety mindset and the security mindset
In addition to AI infosec... |
972c8fe9-e2a2-4a54-a396-f28badeb4a0d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | How does AI Risk Affect the Simulation Hypothesis?
*Preface: I know this post is pretty wild. Just keep in mind that it’s meant to be lighthearted speculation. I hope that LW will eventually have a subforum for this type of post. Until then, please accept my apologies for putting it here.*
Some people (myself incl... |
cba0f462-3ce3-4181-852b-a5916c535679 | trentmkelly/LessWrong-43k | LessWrong | Debate Minus Factored Cognition
AI safety via debate has been, so far, associated with Factored Cognition. There are good reasons for this. For one thing, Factored Cognition gives us a potential gold standard for amplification -- what it means to give very, very good answers to questions. Namely, HCH. To the extent th... |
648a2aac-31cf-47be-b167-d97a36a9e39c | trentmkelly/LessWrong-43k | LessWrong | The Golden Age of Data
The more recent something is, the more data we have about it.
This is true in all fields where collecting information over time is important, including astronomy, archaeology, geology, paleontology, the internet, and even our own memories. Why is this so universal? It's not as if 100-year-old ... |
0e957ede-982c-49b4-93e9-b4111c28f165 | trentmkelly/LessWrong-43k | LessWrong | Counter-irrationality
I don't think anyone here is perfect, though a lot here is simply brilliant logic. Though, as humans, we're automatically biased for or against something, at times for the wrong things and reasons. No bias at all isn't something we're equipped for, and to me sounds quite dull. Some things we have... |
9df3ad15-14b0-46eb-8169-0a07fdfcc0fd | trentmkelly/LessWrong-43k | LessWrong | Nothing Is Ever Taught Correctly
I find that almost nothing, especially not math, is taught from "first" "principles". To gain true understanding of a thing, you must do three things:
First, reduce it to a set of definitions and/or instruction steps; a program.
Second, explain what this program relevantly achieves... |
3677f668-a6a3-4eac-8765-3810e9435f82 | trentmkelly/LessWrong-43k | LessWrong | Trusting Expert Consensus
I recently read Bryan Caplan's The Myth Of The Rational Voter. It's an excellent book in a lot of ways, and one of those ways is how it got me thinking about the issue of expert consensus. It's pushed me more towards thinking that hard data on what the experts in a given field believe about t... |
ba7f91fa-562e-4e93-a7b6-0f5654c61945 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Brussels meetup
Discussion article for the meetup : Brussels meetup
WHEN: 15 December 2012 01:00:00PM (+0100)
WHERE: Rue des Alexiens 55 1000 Bruxelles
We are once again meeting at 'la fleur en papier doré' close to the Brussels Central station. If you feel like an intelligent discussion and are in the nei... |
9725346e-8094-47e0-be60-29d117557079 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Gathering Strength, Gathering Storms: The One Hundred Year Study on Artificial Intelligence (AI100) 2021 Study Panel Report
1Gathering Strength,
Gathering Storms:
In the five years since we released the first AI100 report,
much has been written about the state of artificial
intelligence and its influences on soci... |
bfdbfa16-cb49-465d-8779-53d01bb87835 | trentmkelly/LessWrong-43k | LessWrong | Kolmogorov Complexity Lays Bare the Soul
Kolmogorov complexity is a measure of the algorithmic complexity of a particular piece of information, and it is found by identifying the shortest possible representation of that information. For example, if I have a string of characters AAAAA, I can create a compressed represe... |
45ed319c-52bf-4a38-91fe-e92a9a50d365 | trentmkelly/LessWrong-43k | LessWrong | Beginning at the Beginning
I can't help but notice that some people are utilizing some very peculiar and idiosyncratic meanings for the word 'rational' in their posts and comments. In many instances, the correctness of rationality is taken for granted; in others, the process of being rational is not only ignored, but... |
8cf041ee-a463-4cd6-955e-2aeac5faa722 | trentmkelly/LessWrong-43k | LessWrong | What is your internet search methodology ?
I feel like the heuristics we use to search random tat on the internet are an incredibly larger part of our lives.
I wouldn't make a claim as extended as "googling skills are on the largest determinants of success in life", but I wouldn't look very suspiciously at someone th... |
830b67ce-2a2e-409f-b735-3582bef840c7 | trentmkelly/LessWrong-43k | LessWrong | We can understand value only through endless self-reflection
Once I was talking with a friend about an upcoming career decision. I said I wanted to do something "valuable" and he said that he recognises value when he sees it. Understanding values has never been easy for me, so I was intrigued by his sentence. I starte... |
6a917508-f892-4241-ac58-8549124f7993 | trentmkelly/LessWrong-43k | LessWrong | Challenge proposal: smallest possible self-hardening backdoor for RLHF
A common scenario for how LLMs could lead to AI catastrophe is if some feature of the network becomes AGI and escapes (with whatever subgoal that specific feature was optimized for, instead of the goal that the network itself was trained on).
I no... |
22364638-2a5a-4b56-abff-5f5758b3dd55 | StampyAI/alignment-research-dataset/blogs | Blogs | Time for AI to cross the human range in English draughts
*Updated 26 Oct 2020*
AI progress in English draughts performance crossed the following ranges in the following times:
| | | | |
| --- | --- | --- | --- |
| **Range** | **Start** | **End** | **Duration (years)** |
| First attempt to beginner level | 19... |
be51063c-9344-4ada-9bda-a778484fa8ab | trentmkelly/LessWrong-43k | LessWrong | Monthly Shorts 6/22
Nifty Computer Things
I have a website! kellerscholl.com is the product of weeks worth of obsessing and fiddling with the little details, and I’m happy with the outcome.
arxiv-utils makes arXiv papers have names instead of numbers. Works smoothly, works everywhere.
I found a very useful map-build... |
e3e30afe-f46d-43a7-b60a-3699d77bfa88 | trentmkelly/LessWrong-43k | LessWrong | Hall, "Ethics for Self-Improving Machines" (2011)
FYI.
J. Storrs Hall (of Beyond AI) has written a new article "Ethics for Self-Improving Machines" for the new Machine Ethics volume from CUP. Sections of it are viewable on Google Books. At a glance, most of the chapters in this volume have been previously published.
... |
31a2a9bb-a9ef-42b6-8455-ca192e410f5f | trentmkelly/LessWrong-43k | LessWrong | Competitive Markets as Distributed Backprop
Requisite background: high school level programming and calculus. Explanation of backprop is included, skim it if you know it. This was originally written as the first half of a post on organizational scaling, but can be read standalone.
BACKPROPAGATION
If you’ve taken a c... |
3bdaf134-34c0-4a4a-a6aa-86b56c61d9ff | trentmkelly/LessWrong-43k | LessWrong | Singularity Institute $100K Challenge Grant / 2009 Donations Reminder
In case you missed the notice at the SIAI Blog, the Singularity Institute's next Challenge Grant is now running - $100,000 of matching funds for all donations until February 28th, 2010. If you want your donation to be tax-deductible in the U.S. for... |
4633c312-6526-4f3a-9189-285f65909dea | trentmkelly/LessWrong-43k | LessWrong | Why a No-Fly-Zone would be the biggest gift to Putin, and why Zelenskyy keeps asking for it [Linkpost and commentary]
[Epistemic status: My best analysis from purely public sources on a constantly changing conflict in an actively adversarial epistemic environment. I have no special knowledge or expertise.]
This is a ... |
93640137-d8df-4e6a-aac6-4d2ee7d2b7c0 | trentmkelly/LessWrong-43k | LessWrong | Cryonics Cost-Benefit Analysis
cross-posted from niplav.site
> Is cryonics worth it, and if yes, should one cryocrastinate (i.e. postpone signing up for cryonics to a later date)? Betteridge's law of headlines only applies partially here: Yes, it is probably worth it (under plausible assumptions $850k for a 20 year o... |
8fe65c01-7283-40bc-961c-21b440652d04 | trentmkelly/LessWrong-43k | LessWrong | Self-defeating conspiracy theorists and their theories
Everyone who's been on the internet for sometime has likely encountered odd theories regarding every sphere of human activity. The ones that attract the most adherents are characteristically concentrated in the political and economic spheres.
Some of these theor... |
688da196-1da1-4e1d-a09e-fd7041b867c0 | trentmkelly/LessWrong-43k | LessWrong | “Is-Ought” is Fraught
I’ve concluded that the “Is-ought” correlation (not the philosophical problem) is fraught after my kerfuffle with how one moves from is-ness to ought-ness. Ought is derived from personal emotions (on either extremes), and influence from social circles.
The seemingly real pressure one feels from ... |
6b941cb5-e45b-476f-99ef-2fcf4882a900 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | 273. Strategizing in Large Language Models
welcome to the 2074 session for
aisafety.com
today we'll be discussing strategizing
large language models by earlier Hampton
presented by me Lee Randall
we won't be covering everything in the
essay because it was a bit too long and
because uh
some of the things weren't especi... |
771cbc22-7929-4491-b615-02b08df038d7 | trentmkelly/LessWrong-43k | LessWrong | Convincing Arguments Aren’t Necessarily Correct – They’re Merely Convincing
I've been studying a lot of finance lately, and it strikes me that it's a field that requires a very high degree of rationality and ability to cut through the noise to get to correct arguments.
What's nice about investing especially, though, ... |
7fd81746-5422-4dc7-b46f-268460d1969d | trentmkelly/LessWrong-43k | LessWrong | There’s no such thing as a tree (phylogenetically)
[Crossposted from Eukaryote Writes Blog.]
So you’ve heard about how fish aren’t a monophyletic group? You’ve heard about carcinization, the process by which ocean arthropods convergently evolve into crabs? You say you get it now? Sit down. Sit down. Shut up. Listen. ... |
2d485386-07cb-4de3-bd3f-19ff014c5386 | trentmkelly/LessWrong-43k | LessWrong | Feature idea: extra info about post author's response to comments.
I'd really like it if the comments had a special indication for whether the post author themselves karma-upvoted a comment and whether they agreement-upvoted a comment. This would be helpful for improving a feeling of conversation between commenters an... |
b79a0429-9206-41b9-8514-924af6fcdbc3 | trentmkelly/LessWrong-43k | LessWrong | Linkpost: ‘Dissolving’ AI Risk – Parameter Uncertainty in AI Future Forecasting
This post argues that Carlsmith's model, and conjunctive AI doom estimates in general, neglect an important statistical consideration that results in an inflated estimate of AI X-risk. |
482db32e-28c0-4439-aff9-7b8277a07617 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | The Waluigi Effect (mega-post)
*Everyone carries a shadow, and the less it is embodied in the individual’s conscious life, the blacker and denser it is. — Carl Jung*
*Acknowlegements: Thanks to Janus and Jozdien for comments.*
Background
==========
In this article, I will present a mechanistic explanation of the Wa... |
4dba5f8b-213b-407e-9748-011ca66f8217 | StampyAI/alignment-research-dataset/special_docs | Other | A Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy
The survey identifies 45 AGI R&D projects spread across 30 countries in 6 continents, many of which are based in major corporations and academic institutions, and some of which are large and heavily funded. Many of the projects are inter... |
d87568b1-e698-4296-8c4c-a9df476eff6f | trentmkelly/LessWrong-43k | LessWrong | Empiricism in Gameplay
This article relates to a game, being developed by Shiny Ogre Games, based on "The Twelves Virtues of Rationality" by Eliezer Yudkowsky.
> The sixth virtue is empiricism. The roots of knowledge are in observation and its fruit is prediction. What tree grows without roots? What tree nourishes us... |
90f7cb96-ec06-4c1c-be67-cab6ba6cba18 | trentmkelly/LessWrong-43k | LessWrong | Price's equation for neural networks
Price's equation is a fundamental equation in genetics, which can be used to predict how traits will change due to evolution. It can be phrased in many ways, but for the current post I will use the following simplified continuous-time variant:
ddtx=gcov(x,f)=(∇gE[x|g])cov(g,g)⋅(∇g... |
d748e33e-b86e-4d9e-b1ca-ec70e685d5d1 | trentmkelly/LessWrong-43k | LessWrong | What is a decision theory as a mathematical object?
If we want a space of all decision theories, what mathematical objects does it contain? For example, if a decision theory is a function, what are its domain and codomain?
The only approach I'm familiar with is to view expected utility maximizing decision theories as... |
e557eee1-bebc-4dee-93e5-cebc3c9f52dc | trentmkelly/LessWrong-43k | LessWrong | Terminology: <something>-ware for ML?
Would it be useful to have a term, analogous to 'hardware', 'software', 'wetware', 'vaporware' etc.[1], which could be used to distinguish learned/discovered components of software, like gradient-trained DNNs, prompt-hacked LLMs, etc?
EDIT 2024-01-04: my current favourites are 'M... |
658ac7a4-8c6a-4dd3-a264-dbc74762175f | trentmkelly/LessWrong-43k | LessWrong | Causal scrubbing: Appendix
* Authors sorted alphabetically.
An appendix to this post.
1 More on Hypotheses
1.1 Example behaviors
As mentioned above, our method allows us to explain quantitatively measured model behavior operationalized as the expectation of a function f on a distribution D.
Note that no part of... |
ae1c819a-be86-486b-ba24-b96ef76cc74b | trentmkelly/LessWrong-43k | LessWrong | My Mental Model of AI Optimist Opinions
Epistemic status: The text below is a sort of strawman of AI optimists, where I took my mental model for how I disagree with rationalist AI optimists and cranked it up to 11. I personally disagree with every sentence below, and I'm posting it here because I'm interested in wheth... |
e7c077a2-f9be-4dce-96c9-e318443a7234 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Why Does AI Lie, and What Can We Do About It?
how do we get AI systems to tell the
truth this video is heavily inspired by
this blog post Link in the description
anything good about this video is copied
from there any mistakes or problems with
it are my own Creations so large
language models are some of our most
advan... |
7c741ed4-d8df-4a3d-a96c-042e60099332 | trentmkelly/LessWrong-43k | LessWrong | Literature On Existential Risk From Atmospheric Contamination?
I'm researching the possibility of global atmospheric contamination as a potential existential risk (say via the deliberate release of some extremely toxic chemical which achieves global dispersion [cf. microplastics for evidence this is possible], etc.), ... |
1c6bd4ce-9c53-41d3-bd95-9c2ada7be22e | trentmkelly/LessWrong-43k | LessWrong | <|endoftext|> is a vanishing text?
Just sharing this observation and readers / researchers might find it useful or worth further testing...
A bug?
<|endoftext|> is a special token for the GPT models along with its counterpart <|startoftext|>. Per OpenAI's tokenizer, <|endoftext|> has seven tokens:
However, the i... |
2456b2bb-7d04-41aa-9438-06e91d577e9b | trentmkelly/LessWrong-43k | LessWrong | Bike Lights are Cheap Enough to Give Away
While in a more remote area bike lights let you see where you're going, in a city there's usually enough light around for that. But you still need lights, so other people see you. [1] Unfortunately, lots of people end up biking without lights: it's easy to forget them, have th... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.